Class switch recombination: a B cell deliberately mutates and cuts its own antibody locus, and the same enzyme causes MYC-IgH translocations

The paper's proposed RNA-editing mechanism was superseded by the DNA-deamination model (E3); the requirement result stands, the mechanistic title does not.

• AID was shown to deaminate DNA, not RNA, establishing the DNA-deamination mechanism. Petersen-Mahrt SK et al. 2002

Small original patient series; subsequent cohorts broadened the phenotype.

• 'AID deficiency' in humans is not one recessive loss-of-function entity, and the recessive phenotype does not localise what AID does at the switch region. Human AID proteins carrying insertions, replacements or truncations in the C-terminal region retain strong somatic-hypermutation activity while almost completely losing class-switch activity, and such alleles act dominantly in patients - so the CSR arm depends on an interaction with a CSR-specific cofactor that is genetically separable from deaminase activity itself. A null-allele syndrome bounds what AID is needed for, not how it works. Ta VT et al. 2003

Heterologous bacterial system; the inference to mammalian B cells was confirmed by later genetics including E6.

• Direct contradiction of the glycosylase step this claim asserts. The UNG inhibitor Ugi did not block DNA cleavage at the IgH locus during class switching even though it blocked UNG binding to DNA and strongly inhibited switching, and catalytically dead UNG mutants that cannot excise uracil nevertheless rescued switching in UNG-/- B cells. The authors conclude UNG acts at a repair step by an unknown mechanism and that the dispensability of uracil removal for cleavage 'requires a reconsideration of the model of DNA deamination by AID'. A bacterial mutator assay shows AID CAN deaminate DNA; it does not show that this is the cleavage-generating step in a B cell. Begum NA et al. 2004
• A competing model the E. coli experiment cannot exclude. With an AID-oestrogen-receptor fusion timing AID activation in AID-/- B cells, cycloheximide or puromycin given one hour before activation abolished switching while the same drugs given after activation did not - so an event downstream of AID requires new protein synthesis, which a direct DNA-deamination mechanism does not predict and an RNA-editing mechanism does. This is a minority position (this and the UNG result are both from the Honjo laboratory) and the field's consensus favours DNA deamination; it is recorded because the bacterial assay does not discriminate the two models. Doi T et al. 2003

IL6-transgenic mouse plasmacytoma model; the requirement in spontaneous human Burkitt lymphoma is inferred by analogy, not demonstrated genetically.

• A second laboratory measured the same event and found the opposite. Translocations between the Igh switch region and Myc were measured in AID-deficient mice, and AID made no measurable contribution to the generation of the INITIAL translocations; what depended on AID was the outgrowth of translocation-positive cells. On that reading the IL6-transgenic plasmacytoma requirement is a requirement for tumour progression rather than for the breakage-and-joining event, with intrinsic fragility of the switch regions supplying the break. The disagreement between the two groups is on record and unresolved. Unniraman S et al. 2004

Mouse B cells engineered to lack classical NHEJ components; the relative flux through alt-EJ in a wild-type cell is smaller than these genetics imply.

• Undercuts the inference from junction structure to repair pathway, which this graph makes repeatedly at KMT2A junctions. B cells lacking both Ku70 and ligase 4 - classical NHEJ completely removed - still carry out class switching at levels similar to Ku-deficient cells, but unlike XRCC4- or Lig4-deficient cells they generate SUBSTANTIAL numbers of direct, no-homology joins. The authors conclude that more than one form of alternative end-joining operates. Microhomology is therefore not a reliable signature of alt-EJ, and a blunt junction does not establish classical NHEJ. Boboila C et al. 2010

Mouse genetics; 'not required' rests on the absence of a significant reduction in translocation frequency in knockouts, which has limited power for partial effects.

• 'Not required' and 'no role' are different findings, and a translocation-frequency assay cannot tell them apart for a factor that SUPPRESSES the event. In B cells lacking both Ku70 and ligase 4, internal switch deletions are greatly increased and IgH chromosomal translocations to c-myc are augmented - classical NHEJ actively suppresses the very translocation for which Ku80 is scored dispensable here. A knockout that raises translocation frequency and a knockout that leaves it unchanged both fail a 'required?' test, so the genetic separability asserted here is weaker than it reads. Boboila C et al. 2010

Requires an engineered bait break, so the translocation landscape is measured relative to one chosen locus and is not an unbiased map of all rearrangements.

• The association with transcription is confounded by three-dimensional genome organisation, which this assay does not control. Combining Hi-C of G1-arrested pro-B cells with the same genome-wide translocation sequencing, and normalising genomic DSBs with ionising radiation to remove the break-frequency term, translocations were enriched in cis along the bait chromosome and across other chromosomes in direct proportion to PRE-EXISTING spatial proximity. Transcribed loci are also the spatially clustered ones, so 'concentrated at transcribed loci' and 'concentrated at loci that sit near the bait in the nucleus' are not separated by this experiment. Zhang Y et al. 2012

Same bait-break limitation as E7; published back-to-back with it and should be read as a pair.

• Bounds the word 'independently'. Measuring nuclear contact frequency and DNA-damage rate simultaneously in cultured mouse B cells showed that, in the absence of recurrent DNA damage, translocations between Igh or Myc and every other gene are directly related to their contact frequency; only where recurrent site-directed damage exists does break rate govern the landscape. A bait-break capture method therefore measures contact frequency with the bait as much as it measures AID targeting, and the two capture methods share that architecture rather than being independent of it. Hakim O et al. 2012